Analyzing Heart Rate Variability for COVID-19 ICU Mortality Prediction Using Continuous Signal Processing Techniques.

Area of Science:

• Critical Care Medicine
• Cardiology
• Data Science

Background:

• Heart Rate Variability (HRV) is explored for predicting disease and mortality, but optimal features remain undefined.
• The COVID-19 pandemic highlights the need for early mortality prediction in critically ill patients.
• ECG signals from ICU admission offer a potential window for HRV analysis.

Purpose of the Study:

• To investigate the potential of HRV analysis for early prediction of in-hospital mortality in COVID-19 ICU patients.
• To identify optimal HRV features and analytical methods for risk stratification.

Main Methods:

• Retrospective observational study analyzing ECG signals from 82 COVID-19 ICU patients.
• HRV indices extracted using sliding windows across various observation intervals.
• Applied feature selection, reduction techniques, and classification models (including LDA, Gradient Boosting, Random Forest).

Main Results:

• Compiling feature means across patient windows (Method D) yielded the best predictive performance.
• Linear Discriminant Analysis (LDA) showed consistent and robust performance, achieving an AUC of 0.82±0.13.
• Other models like Gradient Boosting and Random Forest also demonstrated high predictive accuracy.

Conclusions:

• Short-term HRV analysis is a feasible and clinically relevant noninvasive tool for early risk stratification in critical care.
• This data-driven approach can guide timely therapeutic decisions for high-risk ICU patients.
• Potential to reduce in-hospital mortality through proactive patient management.